Page 27 - ITU Journal Future and evolving technologies Volume 2 (2021), Issue 6 – Wireless communication systems in beyond 5G era

ITU Journal on Future and Evolving Technologies, Volume 2 (2021), Issue 6

data rates to 1 Tbit s −1 is something that end users will for compression techniques, which can avoid the need for
surely welcome but not at an augmented price, since it is the envisioned transmission rates.
not a prominent need. An increase in data rate without 6G Regarding latency, end‐to‐end latency consists of the sum
would still be possible, since, as discussed in [87], 5G has of various contributions [91], [92]. The propagation la‑
already foreseen the inclusion of higher frequencies up tency is the physical distance between communication en‐
to 60 GHz. Thus, covering higher frequencies would not tities. This is physically bounded by the speed of light and
necessarily require the huge additional investments for a can only be reduced via techniques that somehow ’vir‐
new 6th generation in order to get additional very broad tually’ shorten the real communication distance such as
bandwidth with frequencies reaching 1 −3 THz. Further‐ MEC. Next, transmission latency depends on the inverse
more, the targeted data rate of 5G set to 1 −10 Gbit s −1 of the available capacity on the communication link. The
can be enough for the satisfaction of many possible ver‐ queuing latency comes from the scheduling of data trans‐
ticals already envisioned or to come. For example, the missions (e.g. prioritisation) and routing. In a network
initial requirements estimated for 3D holographic com‐ ecosystem, where softwarization makes in‐network com‐
munications [40], and implicitly for the digital twin, of puting the pillar of any network functionality, the pro‐
4.32 Tbit s −1 are for raw data. This quantity could signi i‐ cessing is a key aspect. Thus, the processing latency is
cantly be reduced with advanced data compression tech‐ the delay, which depends on the hardware’s processing
niques that could be researched. This means that more capacity of network nodes.
icient and effective methodologies to compress data
could avoid the usage of unnecessary spectrum and un‐ By considering that 6G will be a fully‐intelligent network,
necessary investments by the operators. In this direction, another latency variable jumps into the calculation of
recent works have been exploring the potential of Seman‑ end‐to‐end latency. We may call it intelligence latency. By
tic Communications [88] to achieve maximum compres‐ considering the complete deployment and integration of
sion of data while ensuring the correct accomplishment AI into the communication network management and op‐
of identi ied tasks between interacting entities. erations, it is important to notice that a new paradigm
arises, also bringing with it its own cons: big data. In fact,
Moreover, since the advent of 4G services, network traf ic
intelligence requires continuous big data collection, pre‐
has been drifting from mainly downlink to uplink intense processing and analysis performed in a distributed man‐
usage, because of various new bandwidth‐intensive appli‐
ner by various data centres within the whole network.
cations [89]. Representative examples of services driving
This means an explosion of the control traf ic, which can
to such a paradigm shift in cellular networks are video
become comparable to the amount of data traf ic sent
sharing, real‐time MEC of loading support, cloud backup,
across the network. This means that a 6G intelligent con‐
massive IoT data gathering, etc. With 5G we are already
trol plane [93] will require time not only for distributed
experiencing such dramatic inversion of the direction of
data mining and classi ication but also for training, decid‐
ic. With 6G, many of the new types of services will
ing, and acting on the network environment (we do not
even push the network usage to higher imbalances, with consider now the delay for the network recon iguration
much more uplink traf ic in many scenarios and use cases.
after the intelligent algorithm has acted). In this context,
the use of proactive ML algorithms may help to make la‐
Each UE connected to the network implies a speci ic
amount of processing load required at the BBU. The pro‐ tency ultra‐low [94] or ’negative’ [95]. This paradigm is
cessing required by a UE for the uplink in 4G was ex‐ the so‐called ’anticipatory networking’. The two terms of
pressed in [90] as this paradigm [96] include anticipation, the exploitation
of prediction techniques or the assumption of given fu‐
1 ture knowledge, and networking, the optimisation of mo‐
2
= (3 + + ) 10 (1) bile communications. However, anticipatory networking
3
is somehow incompatible with targeting null failure be‐
where is the number of antennas, the modulation cause prediction (and so ML algorithms) is not determin‐
bits, the code rate, the number of spatial MIMO‐layers istic and it has a variable accuracy (with probability al‐
and the number of Physical Resource Block (PRB). The ways less than 1), whose quality also strictly depends on
processing load is measured in Giga‐Operations Per the data previously collected.
Second (GOPS). Even if de ined for 4G, Eq. ((1)) can give
us an idea of the impact that requirements and trends of Finally, softwarization implies the deployment of virtual
6G previously mentioned can have on computing, latency, machines and containers. It is well‐known that the com‐
and energy usage. A data rate in the order of Tbit s −1 plete and massive use of software is not as ef icient as
increases the value of as the number of MIMO anten‐ the hardware‐based solutions [1], [97]. This means that
nas that 6G is planning to use augments . For example, the open challenge of matching 5G latency requirements
in [73], it is envisioned the employment of 1024×1024 with softwarization will be physically quite hard, espe‐
MIMO elements. This means that the data rate will highly cially when targeting the extreme range of values less
increase computing, which will increase the overall la‐ than 1 ms with 6G. Additionally, the explosion of the data
tency and energy usage. This concern underlines the need rate to 1 Tbit s −1 will probably make impossible the re‐

© International Telecommunication Union, 2021 15